Collaborative/static subagent profiles with configurable models

Resolved 💬 2 comments Opened Feb 4, 2026 by vjain419 Closed May 10, 2026

Feature request

Add a collaboration feature to create static subagent profiles with explicit configuration (e.g., model choice), so different subagents can be assigned to different roles like planning, investigation, and execution.

Problem

Today subagents are ephemeral and not easily configurable by role. Users often want higher‑intelligence models for planning/investigation, but a lightweight model for coding and running jobs to save cost/time. Without role‑based configuration, this workflow is hard to implement consistently.

Proposed solution

Introduce a way to define persistent/"static" subagent profiles, for example:

  • planner (high‑capability model, long context)
  • investigator (high‑capability model for research/analysis)
  • executor (lightweight model for coding/runs)

The user should be able to:

  • create/update/delete named subagent profiles
  • set model, context budget, tools access, and limits per profile
  • assign a profile when spawning a subagent
  • optionally set defaults per task type

Acceptance criteria

  • Users can define named subagent profiles with model + constraints
  • Subagents can be spawned with a specific profile
  • Profiles persist across sessions/projects (or at least per repo)
  • Clear UI/CLI affordance to list and edit profiles

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗